A dashboard that tells the truth

The request#

A dashboard to log lottery draws by date (TOTO: 6 numbers from 1–49 plus an additional; 4D: 23 four-digit numbers per draw), store everything in one Excel file, and suggest "high probability" numbers.

The honest part first#

So instead of pretending, the dashboard measures. It includes a backtest: before each saved draw, it asks each strategy to pick numbers using only the earlier draws, then counts how many matched. It compares that with what random picks would average:

• TOTO, 7 picks from 49 with 6 drawn: random gets 7 × 6 ÷ 49 ≈ 0.86 matches per draw.
• If "Hot" averages about 0.86 too, the data itself shows it's no better than random.

Architecture: why a tiny server?#

A web page in a browser can't write to an Excel file on your phone. So:

index.html (dashboard in the browser)
     │ fetch("api/draws", ...)
     ▼
server.py (Python, runs on the phone, port 8765)
     │ openpyxl
     ▼
Millionaire.xlsx  ── sheet "ToTo" and sheet "4D"

• openpyxl reads and writes Excel files without Excel installed.
• One workbook, one sheet per game; a shared helper adds sheets safely.
• 4D numbers are stored as text, otherwise Excel turns 0427 into 427.

Validation: the boring hero#

Before saving a TOTO draw: exactly 6 main numbers, all 1–49, no repeats, an additional number not among them, and no second draw for the same date unless "Replace" is ticked. Each failure has a plain message ("Main numbers must all be different.").

Testing on a copy#

Every test ran against a copy of the Excel file in a scratch folder, so real data was never touched by experiments.

Key takeaways#

• When a request rests on a false premise, say so kindly, then build something genuinely useful.
• Backtesting against a random baseline is how you check any "strategy".
• Browsers can't write local files; a small local server can.
• Validate input strictly, test on copies.

Quick quiz#

Try it yourself#